Role of Apolipoprotein A-I in the Structure of Human Serum High Density Lipoproteins

For a better definition of the role of human serum apolipoprotein A-I (apo A-I) in high density lipoprotein structure, a systematic investigation was carried out on factors influencing the in vitro association of this apoprotein with lipids obtained from the parent high density lipoprotein (HDL); these lipids include phospholipids, free cholesterol, cholesteryl esters, and triglycerides. Following equilibration, mixtures of apo A-I and lipids in varying stoichiometric amounts were fractionated by sequential flotation, CsCl density gradient ultracentrifugation, or gel-permeation chromatography, and the isolated complexes were characterized by physicochemical means. As defined by operational criteria (flotation at density 1,063 to 1.21 g/ml), only two types of HDL complexes were reassembled; one, reconstituted HDLS, small with a radius of 31 A, and the other, reconstituted HDLL, large with a radius of 39 A. The two types incorporated all of the lipid constituents of native HDL and contained 2 and 3 mol of apo A-I, respectively. A maximal yield of reconstituted HDL (R-HDL) was observed at an initial protein concentration of 0.1 muM, where apo A-I is predominantly monomeric. At increasing protein concentrations, the amount of apo A-I recovered in R-HDL was found to be proportional to the initial concentration of monomer and dimer in solution. The composition and yield of the complexes were independent of ionic strength and pH within the ranges studied. Both simple incubation and cosonication of apo A-I with HDL phospholipids produced complexes of identical composition, although the yeild of complexes was higher with co-sonication. When the comparison of the same methods was extended to mixtures of apo A-I and whole HDL lipids, the results confirmed previous observations that co-sonication is essential for the incorporation of the neutral lipid into the R-HDL complexes. The results indicate that (a) in vitro complexation of apo A-I with lipids is under kinetic control; (b) apo A-I can generate a lipid-protein complex with properties similar to those of the parent lipoprotein; (c) the process requires well defined experimental conditions and, most importantly, the presence in solution of monomers and dimers of apo A-I; (d) the number of apo A-I molecules incorporated into R-HDL determines the size and structure of the reassembled particle. All of these observations strongly support the essential role of apo A-I in the structure of human HDL.

For a better definition of the role of human serum apolipoprotein A-I (ape A-I) in high density lipoprotein structure, a systematic investigation was carried out on factors influencing the in vitro association of this apoprotein with lipids obtained from the parent high density lipoprotein (HDL); these lipids include phospholipids, free cholesterol, cholesteryl esters, and triglycerides.
Following equilibration, mixtures of apo A-I and lipids in varying stoichiometric amounts were fractionated by sequential flotation, CsCl density gradient ultracentrifugation, or gel-permeation chromatography, and the isolated complexes were characterized by physicochemical means. As defined by operational criteria (flotation at density 1.063 to 1.21 g/ml), only two types of HDL complexes were reassembled: one, reconstituted HDL,, small with a radius of 31 A, and the other, reconstituted HDL,, large with a radius of 39 A. The two types incorporated all of the lipid constituents of native HDL and contained 2 and 3 mol of apo A-I, respectively. A maximal yield of reconstituted HDL (R-HDL) was observed at an initial protein concentration of 0.1 PM, where apo A-I is predominantly monomeric. At increasing protein concentrations, the amount of apo A-I recovered in R-HDL was found to be proportional to the initial concentration of monomer and dimer in solution. The composition and yield of the complexes were independent of ionic strength and pH within the ranges studied. Both simple incubation and cosonication of apo A-I with HDL phospholipids produced complexes of identical composition, although the yield of complexes was higher with co-sonication.
When the comparison of the same methods was extended to mixtures of apo A-I and whole HDL lipids, the results confirmed previous observations that co-sonication is essential for the incorporation of the neutral lipid into the R-HDL complexes.
The results indicate that (a) in vitro complexation of apo A-I with lipids is under kinetic control; (b) apo A-I can generate a lipid.protein complex with properties similar to those of the parent lipoprotein; (c) the process requires well defined experimental conditions and, most importantly, the presence in solution of monomers and dimers of apo A-I; (d) the number of apo A-I molecules incorporated into R-HDL determines the size and structure of the reassembled particle. All of these observations strongly support the essential role of apo A-I in the structure of human HDL. Apolipoprotein A-I is the most abundant protein constituent of the human serum high density lipoprotein class. Thus, a definition of its structural role in the native complex would be of great interest.
The lipid-binding capability of apo A-1' has been examined in a number of laboratories (1,2). The reported results have differed considerably as to the lipid-binding capacity of this apoprotein.
In some cases, apo A-I was capable of binding only a small quantity of lipids (3)(4)(5); in other instances, it appeared to complex with relatively large amounts (6)(7)(8). Since the influence of protein concentration and medium conditions was not examined in these studies, the discrepancy in the reported results could be attributable to differences in experimental conditions, especially in concentrations. This possibility was suggested by previous observations from this (9) and other laboratories (10,11)  in the isolation of large amounts of the lipid. protein complex. After ultracentrifugation of the lipid/protein mixtures at a density of 1.063 g/ml as described above, the top 1 ml was collected for analysis (see below), and the bottom 2.3 ml was adjusted to a solution density of 1.21 g/ml with NaBr and centrifuged for 24 h at 11" in a 40.3 rotor at 38,000 rpm. The top 1 ml containing reassembled HDL particles was collected and analyzed (see below). This technique was also employed in the isolation of a lipid-rich particle.
This particle, which floated at density 1.063 g/ml, was separated from free lipid by adjustment of the density to 1.006 g/ml and by subsequent centrifugation for 24 h at 11" in a 40.3 rotor at 38,000 rpm. The bottom 2.0 ml containing the lipid-rich particle was collected and analyzed (see below  (27).

Immunological Studies
Ouchterlony double diffusion experiments were carried out in 1% agarose in 0.05 M Verona1 buffer, pH 8.6, at 4". The immunoprecipitin lines were stained with Amido black after antigen or antibody excess had been washed off with 0.15 M NaCl.
Rabbit antisera against human HDL and apo A-I were prepared as described previously (14).

Analytical Ultracentrifugation
For flotation analyses, we used a Beckman model E analytical ultracentrifuge equipped with electronic temperature and speed controls and a photometric scanning optical system. The reassembled HDL particle isolated by preparative ultracentrifugation was first dialyzed extensively against an NaCliNaBr solution of density 1.21 g/ml and floated at this density. The analyses were conducted at 20", 44,000 rpm, with an AN-H rotor and a double sector cell. The reciprocal of the apparent sedimentation coefficients obtained at various dilutions was plotted against protein concentration to yield the corrected sedimentation coefficient ls~,.,,,]. The lipid-rich particle was analyzed at density 1.063 g/ml, after removal by ultracentrifugation, of the material floating at density 1.006 g/ml.  Fig. 1. The lipid-free apo A-I banded at a density of 1.28 g/ml (Fig. 1C) in the same position as that occupied by apo A-I in the absence of lipids (Fig. 1A). A band of proteinfree lipid which floated to the top of the tube (Fig. 1, C andD) was identical with that of lipids alone (Fig. 1B). The lipid'apo A-I complex had a hydrated peak density of 1.13 g/ml. This complex was similar to that of native HDLa in the broadness of its elution profile and, thus, behaved as a discrete particle with a distinct lipid and protein composition (Fig. 1, C andD). Varying the ionic strength of the medium between 0.5 mM and 0.5 M, or the pH between 6.6 and 8.6, had no effect on the yield or on the lipid/protein composition of the lipid. apo A-I complex.
The banding hydrated density of the lipid.protein complex formed varied with the initial lipid/ape A-I weight ratio (compare Fig. 1, C and D, with Fig. 3; also see Table  VI). In all instances studied, however, the lipid. apo A-I complex floated at a density of 1.063 to 1.21 g/ml; thus it was similar to native HDL in operational terms.
We will refer to this reassembled complex as R-HDL, using the subscript S (small) or L (large) to refer to the radius of the R-HDL particle ( lipids at an initial lipid/protein weight ratio of 1.
amounts by the quick and convenient method of ultracentrifugal flotation. Utilizing this method, we observed no differences in the yield or in the lipid/protein composition of the R-HDL particles compared to those obtained after separation by CsCl density gradient ultracentrifugation.
The fractionation of the sonicated lipid/ape A-I mixtures by agarose column chromatography also afforded a good separation of the reacted and unreacted components. The lipid-free protein and the free lipids eluted near the inclusion volume and the void volume of the column, respectively, whereas the lipid. protein complex eluted as a broad peak between these two peaks.
To establish conditions for maximal complexation between HDL lipids and apo A-I, we examined the effect of incubation versus co-sonication on the interaction of apo A-I with either HDL phospholipids or the whole of the HDL lipids (Table I). With mixtures of apo A-I and HDL phospholipids, no differences were found in the composition of the complex formed either by incubation or co-sonication, although the yields were higher with co-sonication. When apo a-1, in turn, interacted with the total HDL lipids, co-sonication not only produced an actual increase of apo A-I and phospholipid incorporated into the lipid. protein complex, but also promoted the incorporation of both unesterified and esterified cholesterol. Under all conditions, the specific activity of the cholesteryl esters was the same as that of the starting material, indicating that the In addition, the apo A-I recovered in the lipid. protein complex after sonication, when subjected to a quantitative double antibody radioimmunoassay analysis (351, was indistinguishable from native apo A-I. Effect of Apo A-Z Concentration -To determine the influence of protein concentration on yield and on the lipid-binding capacity of apo A-I, we next studied the reassembly of apo A-I at different apo A-I concentrations, but at a constant ratio of protein to lipid. As shown in Table II (Table II; (Table II). The above experiments suggested a possible relationship between the state of association of the protein and its lipidbinding capacity. To investigate further the dependence of the lipid-binding capacity of apo A-I on protein oligomerization, we first determined the state of aggregation of apo A-I under our experimental conditions. For this purpose, we passed lipidfree apo A-I, in varying concentrations, through agarose columns calibrated with proteins of known molecular weight and estimated the apparent molecular weight of the eluted material. At low concentrations (0.1 PM), apo A-I eluted predominantly as a monomer ( Fig. 2A), whereas oligomer forms (dimers and tetramers) prevailed at higher protein concentrations (1 PM, Fig. 2B, and 6.5 PM, Fig. 2C).
Previous work in this laboratory by frontal elution-gel filtration chromatography (9) showed that the multiple species of apo A-I in solution attain equilibrium rather rapidly at 20". Our gel filtration studies suggest, however, that the establishment of the oligomerization equilibrium at 8" is much slower than the time required for the experiment. Indeed, monomeroligomer peaks were symmetrical and clearly resolved as shown in Fig. 2. In addition, the amount of each oligomeric species seemed to be independent of the time required for passage of the solution through the column. For these reasons, one can assume as a first approximation that, at 8", little interconversion of the oligomeric forms of apo A-I occurred during gel filtration. With this assumption, the self-association of apo A-I can then be assessed from our gel filtration studies. The concentration of each oligomeric species, as determined by graphical integration from the gel filtration-elution diagram (Fig. 2), is shown in Table III (Columns 2 to 4) for various initial apo A-I concentrations.
Column 5 of Table III lists the concentrations at which apo A-I is incorporated into the complex, as calculated from the results in the second row of Table II. From these data, it is evident that the amount of apo A-I incorporated in R-HDL, correlated best with the sum of the amount of this apoprotein present in solution as monomer and dimer (Column 4); only at an initial apo A-I concentration of 0.1 pM was there a good correlation with the monomer of apo A-I in solution. At higher concentrations, no Interaction of Apo A-I with HDL Lipids  The solutions of apo A-I of varying concentrations were equilibrated in 4 ml of 0.02 M EDTA, pH 8.6, for 2 h at 4", and then cosonicated with HDL lipids, when present, at an initial lipid/protein weight ratio of 1 in a total volume of 6.4 ml under "Materials and Methods." The sonicated mixtures were centrifuged at 11" for 24 h at a density of 1.006 g/ml, followed by CsCl density gradient ultracentrifugation (see "Materials and Methods"). Lipid-free apo A-I and the lipid. apo A-I complex are the fractions shown in Fig. 1 in 0.02 M EDTA, pH 8.6 (4 ml, A andB; 6.2 ml, Cl, for 4 h at 4", centrifuged, and followed by 8% agarose column chromatography as described under "Materials and Methods." V, refers to the column void volume and V,, to the total column volume (void plus internal volume). A, 12 pg or 0.1 PM apo A-I; B, 112 pg or 1 PM apo A-I; C, 1200 pg or 6.5 PM apo A-I. The data in the subheading and from the second row of Table II  are the molar concentrations in the 1st and 5th columns of this table. The data in the 2nd and 3rd columns were calculated from elution profiles of apo A-I on agarose columns at each specified concentration. The area corresponding to the dimeric or monomeric forms of apo A-I was divided by the total area of the oligomeric forms of apo A-I, and the quotient was multiplied by the amount of apo A-I in the initial mixture.
The data in the 4th column give the sum of data in Columns 2 and 3. The conditions for the fractionation of apo A-I by agarose column chromatography were those described in the legend of Fig. 2. The molarity of apo A-I is expressed in terms of the molecular weiaht of the monomer. weight ratio of 1, R-HDLs had a peak hydrated density of 1.13 g/ml and was well resolved from lipid-free apo A-I and proteinfree lipid. Under these experimental conditions, no other lipid.protein complexes were observed. The properties of this complex, R-HDLs,.,B (the subscript 1.13 referring to the hydrated density of the particle), shown in Table IV, Column 2, were similar to those of native human serum HDLB in Table  IV, Column 3, except that the content of cholesteryl esters was significantly lower (see also Table VI). The kinetics of hydrolysis of the R-HDL,,.,, particle by Crotalus adamanteus phospholipase A, were the same as those previously described for native HDL3 (29, 30).

Znteraction
of Apo A-Z with HDL Lipids 1213 Comparison of properties of two R-HDL, particles and native HDL, Apo A-I was co-sonicated at an initial concentration of 6.5 pM at an initial lipid/protein weight ratio of either 0.5 or 1.0 and centrifuged at a density of 1.063 g/ml as described under "Materials and Methods." The density 1.063 g/ml undernatants were then fractionated by CsCl density gradient ultracentrifugation, ultracentrifugal flotation at density 1.21 g/ml, or gel permeation chromatography as described under "Materials and Methods." The properties of R-HDLsm and R-HDL,, 13, formed at an initial lipid/protein weight ratio of 0.5 and 1.0, respectively, are listed in Columns 1 and 2, respectively.
For comparison, properties of native human serum HDL, are given in the last column.
R-HDL, 16 R-HW.,.,, When solutions containing 6.5 PM apo A-I were co-sonicated with HDL lipids at an initial lipid/protein weight ratio less than 1, the hydrated density and lipid composition of the lipid.protein complexes formed were found to differ from those obtained at a weight ratio of 1. For example, at an initial lipid/ protein weight ratio of 0.2, a dominant lipid. protein complex with a peak density of 1.21 g/ml was observed together with a small amount of complex of a peak density of 1.15 g/ml (Fig. 3). Approximately 55% of apo A-I remained lipid-free and peaked at density 1.28 g/ml, and the protein-free lipids banded at the top of the tube. As the initial lipid/protein weight ratio was increased above 0.2, a decrease in the amount of lipid.protein complex of density 1.21 g/ml occurred concomitant with an increase in the amount of a complex with smaller densities. Compared to native HDL3, the lipid. apo A-I complexes which formed below a weight ratio of 1 were impoverished in lipid (see Table VI). One of the R-HDLs particles, R-HDL,,.,, (the subscript 1.16 again referring to the hydrated density of the particle), at an initial lipid/ape A-I weight ratio of 0.5, was partially characterized; its properties are summarized in Table  IV, Column 1.
At initial lipid/ape A-I weight ratios above 1, CsCl density gradient ultracentrifugation poorly resolved a broad component with general properties similar to R-HDL, (see Table VI and contained mostly lipid with some protein. The latter particle will be referred to as the lipid-rich particle. For example, at an initial lipid/protein weight ratio of 4 and an initial apo A-I concentration of 6.5 PM, approximately 30% of apo A-I remained lipid-free, whereas about 20% and 50% were recovered in R-HDL, and the lipid rich particle, respectively. When the initial lipid/protein weight ratio was increased above 6, both R-HDLL and the lipid-free apo A-I disappeared, whereas the lipid-rich particle became the predominant component. Gel permeation chromatographic studies corroborated the density gradient ultracentrifugation observations. As the initial lipid/protein weight ratio of the co-sonicated mixtures increased above 1, a significant increase in the components of higher particle size was obtained. For example, at our initial lipid/protein weight ratio of 4 and an initial apo A-I concentration 6.5 PM, approximately 40% of the eluted apo A-I was lipidfree, and 50% was part of a heterogeneous lipid.protein complex with an apparent particle weight of 15.1 to 17.5 x 104. Agarose columns gave poor resolution of the R-HDL, and lipid-rich particles observed by CsCl density gradient ultracentrifugation.
In addition, a small amount of the lipid-rich particle now eluted in the void volume. The lipid-rich particle, formed at an initial lipid/protein weight ratio of 4 and an initial concentration of 6.5 pM apo A-I, floated at density 1.063 g/ml. Its apparent weight ranged from 17.5 x lo4 to 2.8 x 106, as estimated by gel permeation chromatography. Its weight per cent composition was protein, 11%; phospholipid, 68%; free cholesterol, 9%; cholesteryl esters, 12%. An ultracentrifugal analysis of this lipid-rich component indicated a heterogeneous particle having components of SA,.063j between 2.0 and 48 at 20".
The results of the compositional analyses parallel the observed shift of the lipid. apo A-I complex from a peak density of 1.21 g/ml to smaller densities (see Table VI). In particular, increasing the initial lipidlapo A-I weight ratio entailed a steady increase in the phospholipid content of the particle. In contrast, the content of free cholesterol and cholesteryl esters increased until the lipid/protein weight ratio reached approximately 1; further increases in this ratio did not produce any increase in the cholesterol content of the particle. At the initial lipid/ape A-I weight ratio of 0.2, the limiting factor appears to be the phospholipid content, since all of the initial phospholipid in the reaction mixture was incorporated into the complex. At higher weight ratios, however, the sphingomyelin/phosphatidylcholine ratios increased almost by a factor of 2 (see Table IV), indicating that, when excess phospholipid is present, the complex is preferentially enriched in sphingomyelin. On the other hand, the phospholipid/cholesteryl ester ratio was virtually constant up to an initial lipid/ape A-I weight ratio of 2, whereas the cholesteryl ester/free cholesterol ratio decreased steadily.
Protein incorporation was also studied as a function of the initial lipid/protein weight ratio at a constant concentration of apo A-I (Table V). For this purpose, we determined the distribution of apo A-I between the lipid-free form, R-HDL, and the lipid-rich particle. These data, expressed in molarities at equilibrium, are presented in Columns 2 to 4 of Table V. At initial lipid/ape A-I ratios between 0.2 and 2.0 (Column l), the amount of apo A-I incorporated into the complex was constant within the experimental error, 2.4 to 3.3 FM with a mean ? standard deviation of 2.8 2 0.4 PM (bracketed data, Column 3). These results were in agreement with those obtained under conditions of varying protein concentration at a constant lipid/ protein weight ratio of 1 (compare with bracketed data in Table III, Columns 4 and 5). At lipid/protein weight ratios above 2, a significant decrease occurred in the apo A-I incorporated both in R-HDL and in the lipid-free fraction, accompanied by a countervailing increase of apo A-I in the lipid-rich particle. It is possible that, under these conditions, the amount of R-HDL formed was underestimated, if it was entrapped in the floating lipid-rich particle and had not separated under our experimental conditions. Our results as a whole confirm that the amount of apo A-I recovered in R-HDL depends only on the amount of monomer and dimer present in solution, thus indicating that the generation of R-HDL is not a true equilibrium process.
Data Analyses-The data presented in the previous sections allowed us to answer the question as to how many different classes of R-HDL particles were formed under our conditions of reconstitution, and whether the lipid composition of the particles was a reflection of apoprotein content. To this end, we used the results obtained at varying initial lipid concentration while keeping the initial concentration of apo A-I constant (Table VI). Using the apparent molecular weight and hydrated density obtained from the peak fraction of the lipid. protein complex, we calculated the radius of the individual particles from the relationship M x lOi4 4m-3 d x 6 x 102" = 3 where M = apparent molecular weight of particle, d = hydrated density of particle, and r = radius of particle in A. The value of r is, of course, only operational and is not necessarily a hydrodynamic parameter. From the weight per cent composition of the lipid. protein complex, and with the assumption of 5% triglycerides/particle (values in parentheses, Table  VI), we also calculated the number of molecules/particle for each component. The results, shown in Table VI, indicate that R-HDL can be grouped into two classes: one with a radius of 31 * 2.2 A and 2 mol of apo A-I (R-HDL,), and the other with a radius of 39 ? 1.5 A and 3 mol of apo A-I (R-HDL,,). Thus, the number of apo A-I molecules/particle is the major determinant of the size of the lipoprotein. A slight variation in size exists within each class, probably due to variations in lipid composition. In comparing the 31A and 39 A particles, we note a substantial increase in phospholipid content, from 14 to 40 mol in the 31 A particle and from 55 to 105 mobparticle in the 39 A R-HDL, indicating a good correlation between phospholipid content and particle size within the classes. On the other hand, both cholesteryl esters and free cholesterol increased significantly in the 31 A class, but remained constant in the 39 A class. A comparison of these data with those for native HDL3 indicates that R-HDL,, formed at an initial lipid/protein weight ratio between 1 and 2, contains amounts of protein, phospholipid, and free cholesterol comparable to those of HDL,, but only about half as many cholesteryl esters. The results also show that R-HDL,, formed at an initial lipid/protein weight ratio above 2, contains more phospholipid than expected for an HDL particle. On the other hand, the lipid-rich particle obtained at an initial lipid/protein weight ratio of 4 appears to be not a lipoprotein, but most probably a lipid vesicle with apo A-I adsorbed to it.

DISCUSSION
The present studies have shown that, when lipid-free apo A-I from human serum HDL is re-exposed in vitro to HDL lipids, it can form well defined complexes which are similar in operational terms to a native HDL. The behavior of the reassembled HDL particles is similar to that of native HDL in terms of physical parameters, namely, electrophoresis, ultracentrifugation, and circular dichroism, as well as in enzymatic digestion by snake venom phospholipase AZ. The formation of the two R-HDL particles, one small and containing 2 mol of apo A-I, the other large and containing 3 mol of apo A-I, is kinetically driven and the process does not reach a thermodynamic equilibrium.
Our results have also shown that the reassembly process requires the presence in solution of monomer-dimer forms of apo A-I which, once incorporated into the reassembled particles, determine the size and chemical composition of the latter. A size-limiting role of apo A-I in HDL has been suggested before (3'7) and is validated by the current work. These results confirm preliminary observations made in this laboratory (38) and further support the concept that the state of association of apo A-I has a marked influence on its ability to bind lipids, presumably because the sites involved in protein-protein interactions are also those participating in the association with lipids (9). The correlation between protein concentration and complex formation is in agreement with the results recently obtained in this laboratory, showing that Macacus rhesus apo A-I, which is predominantly monomeric even at high concentrations (up to 1 mg/ml), is incorporated into lipid.protein complexes at a high yield under all conditions (39). We feel that the reason 10.9 (3.0) 9.5 (2.1) 9.8 (2.0)